research papers Acta Cryst. (2006). D62, 1267–1275 doi:10.1107/S0907444906033555 1267 Acta Crystallographica Section D Biological Crystallography ISSN 0907-4449 Application of high-throughput technologies to a structural proteomics-type analysis of Bacillus anthracis K. Au, a N. S. Berrow, a E. Blagova, b I. W. Boucher, b M. P. Boyle, b J. A. Brannigan, b L. G. Carter, a T. Dierks, c G. Folkers, c R. Grenha, b K. Harlos, a R. Kaptein, c A. K. Kalliomaa, b V. M. Levdikov, b C. Meier, a N. Milioti, b O. Moroz, b A. Mu ¨ller, b R. J. Owens, a N. Rzechorzek, b S. Sainsbury, a D. I. Stuart, a T. S. Walter, a D. G. Waterman, b A. J. Wilkinson, b K. S. Wilson, b N. Zaccai, a Robert M. Esnouf a * and Mark J. Fogg a * a Division of Structural Biology, University of Oxford, Wellcome Trust Centre for Human Genetics, Roosevelt Drive, Oxford OX3 7BN, England, b The York Structural Biology Laboratory, Department of Chemistry, University of York, Heslington, York YO10 5YW, England, and c Bijvoet Center for Biomolecular Research, NMR Spectroscopy, Utrecht University Padualaan 8, 3584 CH Utrecht, The Netherlands Correspondence e-mail: [email protected], [email protected]# 2006 International Union of Crystallography Printed in Denmark – all rights reserved A collaborative project between two Structural Proteomics In Europe (SPINE) partner laboratories, York and Oxford, aimed at high-throughput (HTP) structure determination of proteins from Bacillus anthracis, the aetiological agent of anthrax and a biomedically important target, is described. Based upon a target-selection strategy combining ‘low- hanging fruit’ and more challenging targets, this work has contributed to the body of knowledge of B. anthracis , established and developed HTP cloning and expression technologies and tested HTP pipelines. Both centres devel- oped ligation-independent cloning (LIC) and expression systems, employing custom LIC-PCR, Gateway and In-Fusion technologies, used in combination with parallel protein purification and robotic nanolitre crystallization screening. Overall, 42 structures have been solved by X-ray crystallo- graphy, plus two by NMR through collaboration between York and the SPINE partner in Utrecht. Three biologically important protein structures, BA4899, BA1655 and BA3998, involved in tRNA modification, sporulation control and carbohydrate metabolism, respectively, are highlighted. Target analysis by biophysical clustering based on pI and hydropathy has provided useful information for future target-selection strategies. The technological developments and lessons learned from this project are discussed. The success rate of protein expression and structure solution is at least in keeping with that achieved in structural genomics programs. Received 10 March 2006 Accepted 21 August 2006 1. Introduction The pan-European project Structural Proteomics In Europe (SPINE; http://www.spineurope.org/) is primarily aimed at fostering collaborations to yield methods developments, particularly those relevant to high-value eukaryotic, especially human, proteins of biomedical importance. However, two SPINE partners, Oxford and York, have pursued a colla- borative pilot project on a biomedically important prokaryotic target, Bacillus anthracis, as a vehicle for the development of robust technologies for protein expression, crystallization and structure determination. This paper looks at the developments driven by this collaboration and analyses trends that have emerged from the results. The Oxford Protein Production Facility (OPPF) is funded by the UK Medical Research Council to pilot high-throughput (HTP) protein production and crystallization methods. From the outset, its remit has been to tackle challenging biomedi- cally relevant proteins, but as the technical platform for protein production took shape the need emerged to stress-test the pipeline under true HTP conditions. At the same time, the
9
Embed
Application of high throughput technologies to drug substance and drug product development
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Following cloning (pET-YSBLIC), expression and crystal-
lization, the structure of ThiI was solved (Waterman et al.,
2006) in complex with AMP to 2.5 A resolution by MAD
analysis of crystals of a selenomethionine-substituted form of
the protein (Fig. 1a) using data collected on beamline BM14 at
the ESRF, Grenoble. The structure reveals the THUMP fold
to be unrelated to that of previously characterized RNA-
binding domains. In ThiI, the THUMP domain is accompanied
by an N-terminal ferredoxin-like domain. Analysis of
conserved exposed residues indicates that the tRNA-binding
surface is likely to be formed from both domains; however, a
model of the interaction remains elusive. The modified uridine
at position 8 is buried within the core of tRNA in its canonical
L-shaped conformation, implying that structural changes to
the tRNA molecule are necessary to expose it to the PP-loop
pyrophosphatase active site of the enzyme.
Two targets, BA1655 and BA5174, were identified as
members of the Spo0E-like phosphatase family by amino-acid
sequence comparison with the B. subtilis proteins Spo0E,
YnzD and YisI. In B. subtilis, Spo0A is the key regulator of the
sporulation phosphorelay. The threshold concentration of
Spo0A~P required for sporulation is achieved by the activity
of sporulation sensor kinases. Spo0E-like phosphatases
counter this by dephosphorylating Spo0A~P, thereby inhi-
biting sporulation (Ohlsen et al., 1994; Perego et al., 1994).
Whilst crystallization trials with BA1655 and BA5174 failed,
both proteins were considered suitable candidates for struc-
ture determination by NMR, being small (7.5 and 7.9 kDa,
respectively) and highly soluble. The structures were solved in
collaboration with the SPINE partner in Utrecht. Initial15N-HSQC spectra demonstrated a good dispersion typical of
�-helical proteins, with BA1655 exhibiting greater stability
than BA5174, the latter requiring additional temperature and
concentration optimization. Complete sets of NMR spectra
led to structure elucidation of BA1655 (Grenha et al., 2006, in
the press) comprising a dimer in a four-helix bundle and
consisting of two pairs of helices connected by a tight turn
packed in a head-to-tail manner (Fig. 1b), whereas BA5174 is
monomeric and comprises a similar pair of antiparallel
�-helices. These structures indicated that crystallization diffi-
culties probably arose from disorder at the C-terminus. These
tails were too short (13 and 16 residues for BA5174 and
BA1655, respectively) to be reliably predicted as disordered
by the algorithms used during target selection.
3.3. Structures determined by Oxford
A summary of Oxford progress on the first 48 targets and
the current (early) state of progress with the second 78 targets
is provided in Table 1. Overall, 23 structures have been
determined (including complexes with small molecules) for 19
research papers
1270 Au et al. � High-throughput structural proteomics-type analysis Acta Cryst. (2006). D62, 1267–1275
Table 1Progress on different B. anthracis target sets by the York and Oxfordlaboratories.
Work on the first sets of 48 targets for each laboratory is largely finished andgives an indication of potential eventual success rates. Work on the othertarget sets is ongoing.
Slowly growing crystals were obtained from cocrystallizations
with the same substrate and after an initial 2.9 A resolution
data set confirmed the presence of bound substrate and/or
product, a data set extending to 2.3 A resolution was recorded
on beamline ID14EH2 at the ESRF from a cocrystal that took
more than six months to appear and belonged to space group
P212121. This structure was solved by molecular replacement
starting from the substrate-free structure and revealed
d-ribulose-5-phosphate (probably mixed with the product,
d-xylulose-5-phosphate) bound to differing extents in the six
active sites of the hexamer along with catalytic zinc ions (R
factor = 0.211; Rfree = 0.266). These data are the first for a
substrate/product-bound epimerase and allow us to modify
and update the previously suggested model for epimerase
catalysis (Kopp et al., 1999; Jelakovic et al., 2003).
4. Discussion
4.1. Cloning strategies
Traditional methods of cloning using restriction enzymes
require tailoring of the protocol to each individual target and
hence are not amenable to HTP approaches. This has stimu-
lated interest in ligation-independent methods where only the
primers vary between targets. York has developed a custom
LIC-based approach, whereas Oxford has concentrated on
commercial technologies, initially Gateway and more recently
In-Fusion. All methods have proven amenable to parallel-
ization and success rates are high (Table 1) and relatively
independent of target size. Oxford has adapted the cloning
protocols to work effectively on two robotic liquid-handling
stations, the BioRobot 8000 (Qiagen) and the MWG Theonyx
(Aviso GmbH), whereas York are in the process of adapting
their protocols for robotic liquid handling on a Freedom EVO
(Tecan) as part of their newly established High-Throughput
Expression Laboratory (HiTEL). Primer design is also rela-
tively automated; for example, the OPPF primer-design
interface, OPINE, which is built around the Primer3 package,
provides a direct link between the target-selection software
and the LIMS (Albeck et al., 2006). OPINE, along with the
Web Primer resource (http://seq.yeastgenome.org/cgi-bin/
web-primer), was used remotely by York for early primer-
design requirements, although a new primer-design program
developed in-house is now used for all LIC and LIC3C
primers. One advantage of the York and In-Fusion protocols is
that the primers are shorter than those required to produce
the same short-tagged product with Gateway technologies;
furthermore, only one experimental reaction is required,
resulting in significant cost and time savings.
4.2. Expression and purification strategies
Not surprisingly, for bacterial proteins high levels of soluble
expression are routinely observed by York and Oxford using a
research papers
Acta Cryst. (2006). D62, 1267–1275 Au et al. � High-throughput structural proteomics-type analysis 1271
variety of E. coli strains. Purification strategies in both
laboratories are now based on AKTA XPress and Explorer
3D systems (GE Healthcare) and are found to be very effec-
tive. Where the Oxford laboratory requires cleavage of
N-terminal purification tags, this is routinely performed with a
His-tagged 3C protease, which is now seamlessly integrated
into the HTP context and works with high efficiency. The
effect of different tags in expression is discussed in x4.5.
4.3. Crystallization and crystal quality
Oxford has helped pioneer the use of small-volume (100 +
100 nl drop) crystallization trials (Walter et al., 2003, 2005;
Brown et al., 2003) and now has two Cartesian Technologies
MicroSys MIC400 (Genomic Solutions Ltd) robots in heavy
use. York has deployed the Mosquito (TTP Labtech) robot.
Both laboratories have observed an increase in success rate on
reduction of drop volume. Oxford rapidly moved towards a
strategy of optimization using small-volume drops and has
developed standard procedures which have now been imple-
mented on liquid-handling robots and are very effective,
avoiding the need to translate conditions to a larger drop
format (Walter et al., 2005). Crystals from small drops are used
directly for data collection, although these crystals are
frequently too small to be screened using in-house X-ray
facilities. Thus, Oxford has tended to take crystallization trays
to synchrotrons and explore cryopro-
tection strategies on the beamline, an
effective but labour-intensive approach.
In contrast, York has tended to scale up
crystallization-drop volumes during the
optimization process and as a result the
crystals have usually been large enough
for in-house screening; nevertheless,
four of the 21 York structures were
solved using crystals recovered directly
from 96-well screens. The York strategy
allows scaled-up in-house crystals to be
shipped frozen. The effect of the puri-
fication tag on crystal quality is
discussed in x4.5.
4.4. Structure-determination strategies
Most of the targets studied by both
York and Oxford were amenable to
structure determination by molecular
replacement. In general, structure
solutions proceeded in a routine
manner, with automated molecular-
replacement packages, such as CaspR
(Claude et al., 2004) and the CCP4/
eHTPX-supported development of
MrBUMP (Keegan & Winn, manuscript
in preparation; Bahar et al., 2006) often
proving successful. Indeed, several of
the B. anthracis data sets were used to
help in the development and testing of
MrBUMP. Data for one of the Oxford
B. anthracis targets proved particularly
challenging for automated methods and
have helped to drive further develop-
ments. These data (Bahar et al., 2006)
are somewhat unusual in that they are
from a highly mosaic crystal (>2�) that
nevertheless diffracts to high resolution
(1.5 A).
research papers
1272 Au et al. � High-throughput structural proteomics-type analysis Acta Cryst. (2006). D62, 1267–1275
Figure 1Selected structures of B. anthracis proteins solved by the York and Oxford laboratories. (a) Amonomer of ThiI (BA4899) shown as a ribbon representation. The N-terminal ferredoxin-likedomain is coloured green and forms a continuous �-sheet surface with the THUMP domain (red).The enzymatic PP-loop pyrophosphatase domain is coloured blue and shows the position (as a stickrepresentation) of AMP bound at the active site. The glycine-rich linker joining the THUMPdomain to the PP-loop domain is coloured yellow. (b) A ribbon representation of the lowest energystructure (PDB code 2bzb) of the Spo0E-like phosphatase target BA1655 showing the individualmonomers in green and blue. (c) A ribbon diagram showing the hexameric structure of unligandedd-ribulose-5-phosphate epimerase (BA3998). Individual chains are coloured red, orange, yellow,green, light blue and dark blue. The hexameric structure of the protein with bound substrate hasalso been determined.
4.5. The effect of the choice of tag on overall success
Both York and Oxford routinely use vectors that introduce
short His6 purification tags. York has concentrated almost
exclusively on the non-cleavable N-terminal MGSSH-
HHHHH- tag, with a 3C protease-cleavable tag version (pET-
YSBLIC3C) currently undergoing assessment. Oxford
continues to explore cleavable and non-cleavable N-terminal
tags as well as cleavable C-terminal tags since this has been
found to increase the success rates for ‘difficult’ (i.e. non-
bacterial) targets as part of a multi-construct approach
(Siebold et al., 2005). Experience at both laboratories is that
longer tags adversely affect the ability to crystallize. From six
Oxford targets that produced satisfactory levels of soluble
expression with a long tag, only one gave crystals which were
of poor quality while the tag was still attached. After cleavage,
good-quality crystals could be grown for three of these targets.
research papers
Acta Cryst. (2006). D62, 1267–1275 Au et al. � High-throughput structural proteomics-type analysis 1273
Figure 2Scatter plots of grand average of hydropathy (GRAVY; Kyte & Doolittle, 1982) against pI for proteins from B. anthracis. The background shading ofeach plot shows the clusters (A, B and C) identified from an analysis of T. maritima (Canaves et al., 2004). (a) A plot of proteins for all predictedB. anthracis ORFs. (b) A plot of the proteins selected by the York (triangles) and Oxford (circles) laboratories for HTP studies. The largely complete setsof 48 targets are shown as solid shapes, while those for which work is still in progress are shown as outline shapes. (c) A plot of targets for whichsignificant progress has been made at York (triangles) and Oxford (circles). Red (X-ray) and blue (NMR) shapes show structures that have beendetermined; yellow shapes show proteins which have not had structures determined but have yielded crystals; green shapes show proteins which have notgiven crystals but been expressed solubly in sufficient quantities to enter crystallization trials.
In general, it appears that the short N-terminal non-cleavable
tags used extensively by both York and Oxford have not
markedly impaired success rates for crystallization; however, a
quantitative assessment of the effect on crystal quality of the
short tag introduced by the pET-YSBLIC vector is ongoing in
York. Oxford has already generated a cohort of results (N = 13)
to help address the question of whether removing the His tag
is necessary. In four cases new or improved crystals were
obtained after tag cleavage, whilst in only two cases were
crystals better with an uncleaved tag. On the basis of these
results, the Oxford strategy is now to work with cleavable tags
that are as short as possible (both N-terminal 3C cleavable
tags and C-terminal carboxypeptidase cleavable tags). It is our
experience that the presence of the cleavage site does not of
itself significantly reduce protein expression or solubility
compared with a minimal non-cleavable tag.
4.6. Are target-selection strategies justified?
Oxford achieved a success rate of 31% for 48 targets by
repeating efforts and refining protocols, whereas York
achieved a success rate of 21% from the completed set of 48
targets but made fewer attempts at target rescue. Groups of
proteins within the B. anthracis targets showed particular
characteristics. For example, of the 111 York targets entered
into crystallization trials, 19 (17%) represented TIGR-
in purine, pyrimidine, nucleoside and nucleotide metabolism,
the largest single category of proteins in crystallization. All 19
of these targets produced crystals (of varying quality) with
nine leading to structure solution, making this family parti-
cularly crystal-friendly. In contrast, the tRNA synthetases
tackled by Oxford proved extremely difficult to crystallize
despite giving adequate yields of soluble protein and only one
structure (at 2.8 A resolution) was obtained. This experience
reflects the general difficulty of working with tRNA synthe-
tases. As exemplified by the tRNA synthetases, where several
members of a single protein family were tackled in Oxford, at
least one structure was obtained for every family for which
multiple members were targeted (six families)2 supporting the
strategy of selecting paralogues in order to increase overall
success. However, it is interesting to note that the several
conserved hypothetical proteins that were included because
they are widely conserved across bacteria proved difficult. In
fact, only one out of four has yielded a structure to date.
Finally, the B. anthracis targets demonstrate that the use of
disorder-prediction methods, such as RONN and PONDR
(see Esnouf et al., 2006) have utility; an example of this is
provided by work on a conserved hypothetical protein
(BA0541) where redesign of the construct, omitting 18 resi-
dues at the N-terminus, yielded a 3.3 A structure. This struc-
ture was the only hypothetical structure solved, but it is
notable that unlike the other hypothetical proteins, it
belonged to an established fold.
4.7. Comparison with a structural genomics study ofThermotoga maritima
The combined York/Oxford study of proteins from
B. anthracis has been on a much smaller scale than the full-
genome study of T. maritima performed by the JCSG (http://
www.jcsg.org). For that study, 1878 ORFs were cloned and
have so far resulted in 131 structure determinations. An
analysis of the correlates of success with physico-chemical
properties has been made for the T. maritima project
(Canaves et al., 2004) and it appears that some of the findings
apply to our B. anthracis cohort. In particular, when analysed
in terms of the calculated pI and grand average of hydropathy
index, the open reading frames of B. anthracis group in a
similar way to those of T. maritima (Fig. 2a). However, the C
cluster is sparsely populated and the B cluster does not extend
to such high pI values. Although these clusters were not
considered formally in target selection, it can be seen that both
York and Oxford tended strongly to select targets within
clusters A and B (Fig. 2b). In line with the findings of the
JCSG, the targets that led to successful structure determina-
tions were far more likely to come from cluster A (Fig. 2c). It
is intriguing to note that York did manage to solve two
structures from cluster B by NMR analysis. Such plots may
therefore be useful as a direct aid to target selection and
construct design. Overall, the percentage of targets selected by
York or Oxford that yielded structures was higher than that
obtained by the JCSG for their comprehensive attack on
T. maritima, suggesting that the Oxford and York target
selection strategies used were of value in eliminating intract-
able targets.
5. Conclusions
The York and Oxford studies on B. anthracis have yielded
much more than just structures, although the �30% success
rate they have achieved alone makes them very successful
studies. They have provided a test bed for many of the tech-
nological developments in these two laboratories, develop-
ments which have been shared across Europe, with
dissemination largely enabled by the SPINE project. The work
has driven the development of much improved vectors for
ligation-independent cloning strategies, has helped to
optimize expression and purification protocols, has further
validated the HTP nanolitre-scale crystallization methods, has
demonstrated the utility of the target-selection tools devel-
oped at the start of SPINE and has illustrated ways in which
these can be further improved, e.g. by the more systematic use
of disorder-prediction methods and by considering the clusters
in hydropathy and pI. The requirement for conservation
across a series of genomes imposed by the OPPF selection
criteria also appears to have been beneficial in eliminating
difficult proteins, but excludes proteins characteristic of
B. anthracis that may play an important role in pathogenesis.
Structural genomics studies have provided an enormous
impetus for methods development in structural biology and
few laboratories are now untouched by their effects. The
emphasis in SPINE has been to apply these methods to
systems of biological interest, the ultimate aim being to solve
significant problems more effectively by the use of HTP and
parallel technologies. However, the experience with
B. anthracis makes it clear that novel technologies should be
benchmarked on more tractable targets to reveal techno-
logical inadequacies that might otherwise remain hidden
within the high attrition rate observed with more challenging
problems.
We would like to thank Professor Colin Harwood
(Newcastle) for providing genomic DNA. We also thank the
staff of BM14 (Grenoble), the SRS (Daresbury), the ESRF
and EMBL (Grenoble) for access to synchrotrons and help
with data collection. This work was supported by the
European Commission as part of SPINE (Structural Proteo-
mics In Europe) contract No. QLG2-CT-2002-00988 under the
Integrated Programme ‘Quality of Life and Management of
Living Resources’. The OPPF is also funded by the Medical
Research Council UK.
References
AB, E. et al. (2006). Acta Cryst. D62, 1150–1161.Albeck, S. et al. (2006). Acta Cryst. D62, 1184–1195.Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J.
(1990). J. Mol. Biol. 215, 403–410.Alzari, P. M. et al. (2006). Acta Cryst. D62, 1103–1113.Anderson, I. et al. (2005). FEMS Microbiol. Lett. 250, 175–184.Aravind, L. & Koonin, E. V. (2001). Trends Biochem. Sci. 26, 215–217.
research papers
1274 Au et al. � High-throughput structural proteomics-type analysis Acta Cryst. (2006). D62, 1267–1275
Aslanidis, C. & de Jong, P. J. (1990). Nucleic Acids Res. 18, 6069–6074.Bahar, M. et al. (2006). Acta Cryst. D62, 1170–1183.Brown, J. et al. (2003). J. Appl. Cryst. 36, 315–318.Canaves, J. M., Page, R., Wilson, I. A. & Stevens, R. C. (2004). J. Mol.
Biol. 344, 977–991.Cladiere, L., Blagova, E., Levdikov, V. M., Brannigan, J. A., Seror, S. J.
& Wilkinson, A. J. (2004). Acta Cryst. D60, 329–330.Claude, J. B., Suhre, K., Notredame, C., Claverie, J. M. & Abergel, C.
(2004). Nucleic Acids Res. 32, W606–W609.Dean, F. B., Nelson, J. R., Giesler, T. L. & Lasken, R. S. (2001).
Genome Res. 11, 1095–1099.Enright, A. J., van Dongen, S. & Ouzounis, C. A. (2002). Nucleic
Acids Res. 30, 1575–1584.Esnouf, R. M., Hamer, R., Sussman, J. L., Silman, I., Trudgian, D.,
Yang, Z.-R. & Prilusky, J. (2006). Acta Cryst. D62, 1260–1266.Folkers, G. E., van Buuren, B. N. M. & Kaptein, R. (2004). J. Struct.
Funct. Genomics, 5, 119–131.Grenha, R., Rzechorzek, N., de Jong, R. N., AB, E., Diercks, T.,
Truffault, V., Ladds, J. C., Fogg, M. J., Bongiorni, C., Perego, M.,Kaptein, R., Wilson, K. S., Folkers, G. E. & Wilkinson, A. J. (2006).In the press.
Inglesby, T. V., O’Toole, T., Henderson, D. A., Bartlett, J. G., Ascher,M. S., Eitzen, E., Friedlander, A. M., Gerberding, J., Hauer, J.,Hughes, J., McDade, J., Osterholm, M. T., Parker, G., Perl, T. M.,Russell, P. K. & Tonat, K. (2002). J. Am. Med. Assoc. 287, 2236–2252.
Ivanova, N. et al. (2003). Nature (London), 423, 87–91.Jelakovic, S., Kopriva, S., Suss, K. H. & Schulz, G. E. (2003). J. Mol.
Biol. 326, 127–135.Kopp, J., Kopriva, S., Suss, K. H. & Schulz, G. E. (1999). J. Mol. Biol.
287, 761–771.Kunst, F. et al. (1997). Nature (London), 390, 249–256.Kyte, J. & Doolittle, R. F. (1982). J. Mol. Biol. 157, 105–132.Lewis, R. J., Brannigan, J. A., Offen, W. A., Smith, I. & Wilkinson,
A. J. (1998). J. Mol. Biol. 283, 907–912.Lewis, R. J., Scott, D. J., Brannigan, J. A., Ladds, J. C., Cervin, M. A.,
Spiegelman, G. B., Hoggett, J. G., Barak, I. & Wilkinson, A. J.(2002). J. Mol. Biol. 316, 235–245.
Luan, C. H., Qiu, S., Finley, J. B., Carson, M., Gray, R., Huang, W.,Johnson, D., Tsao, J., Reboul, J., Vaglio, P., Hill, D. E., Vidal, M.,DeLucas, L. J. & Luo, M. (2004). Genome Res. 14, 2102–2110.
Manchee, R. J., Broster, M. G., Melling, J., Henstridge, R. M. & Stagg,A. J. (1981). Nature (London), 294, 254–255.
Mock, M. & Fouet, A. (2001). Annu. Rev. Microbiol. 55, 647–672.Ohlsen, K. L., Grimsley, J. K. & Hoch, J. A. (1994). Proc. Natl Acad.
Sci. USA, 91, 1756–1760.Perego, M., Hanstein, C., Welsh, K. M., Djavakhishvili, T., Glaser, P.
& Hoch, J. A. (1994). Cell, 79, 1047–1055.Radnedge, L., Agron, P. G., Hill, K. K., Jackson, P. L., Ticknor, L. O.,
Keim, P. & Andersen, G. L. (2003). Appl. Environ. Microbiol. 69,2755–2764.
Ramabhadran, T. V. & Jagger, J. (1976). Proc. Natl Acad. Sci. USA,73, 59–63.
Read, T. D. et al. (2003). Nature (London), 423, 81–86.Rodrigues, A. & Hubbard, R. E. (2003). Brief. Bioinform. 4, 150–
167.Romero, P., Obradovic, Z., Li, X., Garner, E., Brown, C. & Dunker,
A. K. (2001). Proteins, 42, 38–48.Siebold, C., Berrow, N., Walter, T. S., Harlos, K., Owens, R. J., Stuart,
D. I., Terman, J. R., Kolodkin, A. L., Pasterkamp, R. J. & Jones, E. Y.(2005). Proc. Natl Acad. Sci. USA, 102, 16836–16841.
Studier, F. W. (2005). Protein Expr. Purif. 41, 207–234.Turnbull, P. C. (2002). Curr. Top. Microbiol. Immunol. 271, 1–19.Walter, T. S., Diprose, J., Brown, J., Pickford, M., Owens, R. J., Stuart,
D. I. & Harlos, K. (2003). J. Appl. Cryst. 36, 308–314.Walter, T. S., Diprose, J. M., Mayo, C. J., Siebold, C., Pickford, M. G.,
Carter, L., Sutton, G. C., Berrow, N. S., Brown, J., Berry, I. M.,Stewart-Jones, G. B. E., Grimes, J. M., Stammers, D. K., Esnouf,R. M., Jones, E. Y., Owens, R. J., Stuart, D. I. & Harlos, K. (2005).Acta Cryst. D61, 651–657.
Waterman, D. G., Ortiz-Lombardia, M., Fogg, M. J., Koonin, E. V. &Antson, A. A. (2006). J. Mol. Biol. 356, 97–110.
Yang, Z. R., Thomson, R., McNeil, P. & Esnouf, R. M. (2005).Bioinformatics, 21, 3369–3376.